A STUDENT'S INTRODUCTION TO THE WRIGHT BROTHERS WIND TUNNEL AT MIT1

Eugene E. Covert

T. Wilson Professor Emeritus

INTRODUCTION

A wind tunnel is an instrument whose purpose is to measure some aerodynamic property(ies)

of the body installed in the test section. In some, but not all cases, the body is a sub-scale
model representing an actual structure like an airplane, a building, a racing automobile or
sailboat, etc. In other cases, the body may be a full scale item like a bicycle or a set of
oarlocks, an instrument to be calibrated or configured supporting a research program.

The Wright Brothers Wind Tunnel (WBWT) is a closed return, closed test section wind tunnel.
The test section is basically a 7x10 foot (2.133x 33.281 m) ellipse and is 15 feet (4.877 m) long
(Fig 1). While the tunnel can be pressurized from 0.25 atmospheres to 2.0 atmospheres, an
airlock does not isolate the test section and so testing times at test conditions other than
atmospheric pressure are quite long. So tests at pressurized conditions are rarely conducted,
unless when the testing environment demands such tests.

The tunnel is 'adiabatic' in the sense that there is no active cooling. At temperatures above 90
F, some mechanical elements may bind. Thus testing in the summer is stopped when the
tunnel static temperature exceeds 90 F (32 C). The auxiliary air supply can be used to cool the
tunnel when outside air temperatures are low enough to provide cooling air.

A 2,000 HP alternating current motor powers the WBWT. This motor runs at one of two
constant speeds. The pitch setting of the 6 bladed 13 foot 6 inch diameter fan controls the
airflow speed. Because of a self-imposed noise constraint, the tunnel is usually run at first
speed (up to about 90 mph), and only occasionally in second speed. Thus the upper limit of
speed is about 150 mph (240 kmph). The lowest speed is about 15 mph (24 kmph). Under
these limits the demand power is less than 350 HP.

As implied above, a wind tunnel is an instrument whose purpose is to measure forces and
moments, or the pressure distribution, on models designed for that purpose. The tunnel is
shaped to provide a steady uniform airflow. As an instrument, a wind tunnel consists of several
elements; the wind tunnel shell, the propulsion system including a speed control unit, and a
system of sensors and their associated electronics that record the model orientation with
respect to the air stream, and the forces and moments acting on the model. Additional
recorded data includes both the tunnel total and static pressure and total or static temperature.
These data is used to compute the dynamic pressure, the Mach number, speed of sound, the
tunnel velocity, and the Reynolds number.

1
Rev. 1 July 2004

1
USEFUL HINT

Usually the test conductor needs to use tools to install and to change the model
configuration. The test duration may be materially shortened if the tools are put away in
their proper place after each use. This act reduces the tool search time for the test
conductor and future test conductors.

MODEL CHARACTERISTICS

MODEL SIZE

The maximum model size is dependent upon the type of testing. In the case of aerodynamic
vehicles the model should have a maximum cross section area of about three to five per cent
of the tunnel cross section area, or from about 1.65 square feet to 2.95 square feet. The latter
area should also include the frontal area of the aircraft at the maximum angle of attack. To
reduce the wind tunnel wall upwash at the tip to a negligibly small level, the model's wingspan
should be five feet or less. In any case the wingspan should not exceed six and a half to seven
feet.

Other types of testing, such as aerodynamic loads on buildings, estimation of pedestrian level
winds, and the like, the models can be up to ten per cent blockage, as are many automotive
tests. The test designer for other tests such as those related to athletic performances may
choose larger blockage areas, recognizing the loss of accuracy might be encountered.

MODEL RIGGING

It is important to design the model such that the horizontal axis can be measured when the
model is mounted in the wind tunnel. In addition, it is often desirable to make a pitch run with
the model inverted. (The intersection of a plot of the lift coefficient upright versus angle of
attack, and the inverted model versus the negative angle of attack provides a measure of the
wind tunnel flow angle.) If an inverted run is anticipated, leveling sites will be needed on the
bottom of the model, too.

If angle of roll or angle of yaw are important parameters in your test matrix, be sure your model
design includes a reference points to allow these important angles to be set or measured
accurately.

MODEL DESIGN

As mentioned above the maximum model blocking frontal area should be less than 2.95
square feet. As noted above, while the better choice of wingspan is five feet or less, the
maximum wingspan should exceed not six and a half or so.

In most cases the model should be designed for at least two times the expected maximum

2
static load. A design load of three times the maximum expected load is better than two. In
most cases the maximum pitching or yawing moment is not a critical design parameter; BUT
CHECK IT OUT NUMERICALLY; DON'T JUST ASSUME THIS PART OF THE LOAD CAN BE
NEGLECTED!

To estimate the maximum load, guess at maximum the dynamic pressure, q. To do this recall
that q is about 25.4 psf at 100 miles per hour, and that q changes as the square of ratio of your
velocity to 100 mph.

After the model is constructed it should be loaded to 95% of the maximum design load to
insure the model was constructed as designed.

Finally, if you have several model configurations to test, practice assembling and
disassembling your model prior to you wind tunnel entry.

Nothing will have a greater adverse impact on your schedule than a model failure during test.

PRESSURE MODELS

The primary difference between a pressure model and a force and moment model is the
pressure taps and the channels in the model for pressure lines or wires from imbedded
transducers. If you are planning to conduct a pressure test discuss the location of the static
pressure taps with your advisor. The pressure taps should be flush with the surface, deburred
without rounding the edge of the hole. The pressure tap diameter is usually between 1/16 inch
and at most 1/8 inch for large models. Pressure models are more complicated to build because
of the need to provide a passage for the tubes connecting the pressure taps to the pressure
transducers. Unless the model is very simple, pressure tests require models that are too
complex to be satisfactory student projects.

BALANCE CALIBRATION

In many test programs, the fundamental goal is to determine the forces and moments acting
on a model. These are measured using either a wind tunnel balance installed in the wind
tunnel, or external strain gauge balances, or a special purpose balance made for your test.
The balance is an instrument that resolves the force and moment vectors acting on the model
into forces and moments in a rectangular wind axis coordinate system. In the Wright Brothers
Wind Tunnel lift is usually positive upwards, normal to the wind tunnel longitudinal axis, (more
or less normal to the wind direction), drag is positive aft along the wind tunnel axis (more or
less along the wind direction), and side force is normal to the plane defined by lift and drag and
completes a right handed coordinate system. (On occasion floor mounted half-span wing
models are tested. In this case lift is in the direction of side force). Note the model weight is
negative along the lift axis for non half-span models in the Wright Brothers Wind Tunnel. The
balance system also converts these force and moment components into voltages. (The wind
tunnel data system converts these voltages back into forces and moments.)

3
The process that relates these vector components to output voltages is called calibration. In
this process a special apparatus, called the calibration apparatus, is mounted to the balance at
the point where the force and moment vectors are resolved into components. This point is
called the balance resolving center.

The calibration apparatus enables force and moment components to be applied to the balance
and as such allows the calibrator to define a mathematical model that will convert the voltages
back into forces and moments.

CALIBRATION PROCESS

Usually the calibration process falls in the category of a “Factorial Design of Experiments.” For
each test block, one variable is changed over its entire range, which the others are fixed. For
the next block the initial variable is fixed, and another variable is changed over its range, and
so forth. This leads to a large number of test blocks (actual 6 factorial or 720 blocks). Thus the
process exceedingly tedious and time consuming process. Usually several shortcuts or
truncations are made in the process. These truncations are based upon the assumption the
balance is reasonably well designed and built so many of the interactions can approximated
with far fewer blocks.

A detailed example for one degree of freedom is given below. The example will be a lift
calibration. Suppose the lift loads vary from -50 pounds to plus 150 pounds. For ease of
operation select 20-pound intervals but include zero, too. Be sure to record not only the lift
output voltage, but also the other five components; drag, side force pitching, rolling and yawing
moments. Be sure to note the excitation voltage setting.

The non-lift components are important for precise measurements. They are a measure of the
imperfections in the actual construction and assembly process of the balance. Rarely are these
extra data negligible.

NOTICE

BEFORE APPLYING A LOAD TO THE CALIBRATION APPARATUS; INSURE THE LOAD

CELLS CURRENTLY INSTALLED IN THE BALANCE HAVE A LOAD RANGE THAT
EXCEEDS YOUR CALIBRATION LOAD RANGE

EXAMPLE

Lift Calibration consisting of five loading blocks, -50 to 150 pounds range, with a 20-pound
increment.

Block I Lift load only

Step 1. Record the output for zero applied loads. Record data after the data on the computer

4
 screen is steady, with the exception of a dither in the last place.
Step 2. Carefully locate -50 pounds of lift load on the calibration apparatus. Damp out any
swinging motion by lightly touching the load perpendicular to the load direction. Record
the data after the data n the computer screen becomes steady, except for a dither in the
last significant place. Carefully remove the -50 pounds.
Step 3. Carefully place -30 pounds of weights on the apparatus and then repeat Step 2.
Step 4. Repeat Step 3, using -10 pounds of weights.
Step 5. Repeat Step 3, with zero load.
Step 6. Repeat Step 3 using +10 pounds of weights Steps 7 to 13. Repeat Step 3. using 30,
50, 70, 90, 110, 130 and 150 pounds of weight, respectively.
Steps 14 to 26. Repeat Steps 13 through 1 by removing weights.
Step 27. Repeat Step 1.
Steps 28 to 54.Repeat Steps 1 through 27.

Block II Lift load plus a mean drag load on the drag loading
apparatus. Repeat Steps 1 through 54 above.

Block lII Lift load plus mean drag and mean pitching
moment loads. Repeat Steps 1 through 54 above

Block IV Lift load plus mean pitching

moment load. Repeat Steps 1 through 54
above.

Block V Lift load only Repeat Block I.

Note that this loading plan does not call for a complete set of loads to determine the cross
coupling terms in comparison with a strict Factorial Loading Pattern which requires all
combinations of drag and pitching moment loads. The reduction is practical because the
balance is designed to minimize the cross coupling effects.

Note the calibration of the force and moment balance is the one operation in which you,
the user, control the accuracy of your data. You should strive to keep the standard
deviation of your calibration to less than one part in one thousand. This is essential if
you are aiming at an over all standard deviation of you data between five parts in one
thousand and ten parts in one thousand. Reaching this goal requires care in handling
and hanging the weights.

DATA REDUCTION

Average each load point in Blocks I - V; there will be 12 average load points of 4 readings
each. Call out the overall maximum and minimum, or extreme deviation from all the means of
each block. Compare this maximum deviation with the initial and final zero values for each
Block. The difference between these two readings should not exceed the over all extremes.

If the difference between the first zero reading in Block I and the last zero reading in Block V

5
exceeds the limit defined above, then plot all the values of the zero reading against time.
Linearly curve fit to determine if a sensible drift occurs. If such a drift occurs then interpolate
the average for each loading condition as a function of time and correct them accordingly.

Compare each load point average in Bocks I and V, including the five cross coupling terms, i.e.
drag output for each lift loading, pitching moment output due to lift loading, side force due to lift
loading, etc, with the extreme maximum and minimum described above. If the Block I and
Block V cross couplings lie within the extremes, assume the average for them. These are the
initial values for the cross coupling terms from these two blocks.

Next, fit the best mean square straight line to the average lift output from Blocks I V and
compute the standard deviation. The standard deviation should be less than one Dart in ten to
the fourth.

Then divide cross coupling data in Blocks II, III, and IV by the magnitude of the secondary
load, and compared with the same cross coupling data in Blocks I and V.
If these data fall within the extremes cited above, fit a straight line to the average cross
coupling between the two drag and pitching moment levels.

The first column in a six by six matrix has been defined by the calibration process. If this
process is repeated for the other five components, a 6 by 6 calibration matrix is defined. If the
proposed test is requires measuring six components, three additional blocks, Vl, VII and VlII
will be needed to include an applied rolling moment, side force and yawing moment. These
additional loads slightly enlarge the data reduction process.

In use, the output voltage for each component (a column matrix), is premultipled by the inverse
of the calibration matrix to provide the component force and moment data.

OPERATIONAL PLANNING AND EXECUTION

TEST MATRIX

A “Test Matrix” is really an outline of the steps you plan to take while your test is in process.
Thus the first step in preparing the test matrix is writing your test objective. This is a statement
your goal, or the hypothesis you wish to test, and a description of the data you will need to
accomplish your task. After this statement has been written, the test matrix is prepared.

The test matrix is a table of the step changes for the primary and secondary independent
variables. It is developed from a “Run Schedule.” A truncated run schedule is listed below.

6
 Run Schedule

Run Number Velocity Configuration Angle of Attack Comment

1 50 mph Wing alone -5

2 “ “ 0
3 “ “ 5
4 “ “ 10
5 “ “ 5
6 “ “ 0
{same angles of
7-12 75 mph “
attack
13 -18 100 mph “ as in runs 1-6}
19 - 37 Wing plus flap same as runs 1 - 18

The Test Matrix takes the form below

Sample TEST MATRIX

For each Configuration (Wing alone; Wing with Flap)

Velocity\Angle of Attack (degrees)

50mph 5 0 5 10 5 0
75mph 5 0 5 10 5 0
100mph 5 0 5 10 5 0

(As a wild guess, if all were going smoothly, this sample test would take about a half day.
Usually the first day's accomplishments are substantially less than suggested above).

EXECUTING THE PLAN

During the actual testing process, keeping a test log is ABSOLUTELY ESSENTIAL. The test
log resembles the Run Schedule but without truncation. The test log also includes additional
data like the test room air temperature, the barometric pressure, the tunnel air temperature, the
date and time, as well as configuration, run number and comments. Some of this information is
included in the data stream from the computer, but most test conductors like to have this
information in the log. The comments column is the place to note observations on unexpected
events; like a change in tunnel sound, need for a rerun if the data looks unusual, etc. Keep in
mind the Run Schedule is not a blue print cast in iron, but is a guide. The test log is an
accurate summary of the events as they occur during the test. The test log is invaluable for
resolving ambiguities that are noted when the final report is written.

7
Prior to executing the test, it extremely helpful to carry out some first order approximate
calculations of the data to provide some comfort that the data is reasonable, as the test is
being conducted.

Note that the weight of a model at angle of attack contributes only to the lift load in a wind axis
system, but contributes to both the normal force and axial force in a body axis coordinate
system.

Data Reduction and Presentation

Data is more useful and valuable if it can be expressed in dimensionless form. The choice of
dimensionless variables, and in some cases, the scales on the abscissa and coordinate
should be discussed with your advisor.

Usually the force and moment data are reduced to non-dimensional form by dividing the,
force data by the dynamic pressure, 'q', and the characteristic area,'S'; the pitching moment
data is divided by q, S and a characteristic length, c which is usually the airfoil chord for wings
and for missiles and projectiles the reference length on the maximum body diameter.
If your test involves lift and drag coefficients, there are some recommended standard ~
graphical scales that help the interpretation of the data. For example, when plotting lift l~!
coefficient as a function of angle of attack the length for a change of 8 degrees in angle of
attack should be the same length for a 0.4 change in lift coefficient. Here the angle of attack is
the abscissa and lift coefficient is the ordinate.

The length for a drag polar area 0.004 in drag coefficient is the same as 0.4 in lift coefficient.
Similarly -0.1 in pitching moment coefficient corresponds to 0.4 in lift coefficient. Both these
curves can be plotted on the same graph, with lift coefficient as the abscissa and both the drag
coefficient and the pitching moment coefficient on the ordinate. Some airplane companies
interchange the abscissa and the ordinate.

GENERAL REMARKS ON AIRFLOW

Flow Properties

The flow uniformity is determined by measurement of dynamic pressure and flow angle over
several planes normal to the tunnel centerline. Figure 2 (Ref 1) shows the distribution of the
ratio of the local dynamic pressure to the set pressure in the plane of the balance. The iso-q
ratio lines are shown, together a few recent measurements. This data suggests the velocity
distribution a comparable to newer tunnels Figure 3, (Ref 2) Figure 4 and 5, (Ref 3) shows the
flow angle is acceptably small, but that there is a slight swirl that will appear as an off set on
rolling moment data.

The fundamental aerodynamic property that characterizes any wind tunnel test is the dynamic
pressure. It is one of the key variables in non-dimensionalizing the data and is a defined
quantity.

8
The dynamic pressure q, is defined as
1 (1)
q= ρ ∞V∞2
2
which is identical with
(2)
γ
q = p∞ M ∞2
2
Where

γ −1
 
2 2   pT − p  γ 
M∞ = 1 +
∞
 − 1
γ − 1  p∞  
 

Equation 2 is preferred because it automatically includes the effect of compressibility on q.

The data is sampled at 992 hertz, averaged and printed on the computer screen once a
second. This implies the data is effectively free of aliasing for frequencies of 300 hertz or less.
The wind tunnel speed reaches equilibrium about two minutes after the speed is changed. If
care is taken, the tunnel speed can be reset to about 1 part in a thousand. The bar on the
abscissa on the lower right hand side is the estimate of the standard deviation of the electronic
equipment, Figure 6A. The standard deviation of the measured dynamic pressure as a function
of velocity over commonly used wind tunnel speeds is also shown on Figure 6B. The
instrumentation error is small compared to the wind tunnel fluctuations, which are smaller than
those observed in many wind tunnels. The physical phenomena that gives rise to the peak is
not known. Since no instrument is ideal, the data is distributed about a mean. Figure 6A shows
the one standard deviation of q as a function of tunnel speed. Note this is the absolute value of
standard deviation, not a percentage. The standard deviation of q is three times or greater than
the standard deviation of the instrumentation used to determine q and as such is quite reliable.
The peak value of the standard deviation that occurs around 80 mph, about 0.06 psf, is due to
unknown tunnel phenomena. Figure 6B presents the same standard deviation of q data,
normalized by the local dynamic pressure. This form of the data is useful in estimating the over
all fractional accuracy of the wind tunnel data.

A hot wire anemometer is used to characterize the unsteady or fluctuating velocity level in the
WBWT. When the fluctuations are more or less random, the unsteadiness is called turbulence.
The turbulence level is defined by the root-mean-square velocity fluctuation level divided by
the mean tunnel velocity. In the WBWT the turbulence level is about 0.5 percent, independent
the tunnel speed setting (Ref 4 and 5). This level has a negligible effect on the data for most
tests.

Figures 7, 8, and 9 show the ratio of the turbulent energy to the root mean velocity squared as
a function of the log of frequency for three cases.

The first is with the power on and the fan running at 440 rpm but at zero pitch, so the wind

9
tunnel velocity is zero. The second is for the tunnel running at about 50 mph at the same fan
speed, while the third is at a wind tunnel speed of about 70 mph and a fan speed of 558 rpm.

The 440 rpm peak occurs at log n of 0.865 or at a frequency of 7.33 Hz. The peak at 7.1 Hz is
indistinguishable from the 440 rpm peak, within the accuracy of the data. This frequency
seems match a wave traveling around the circuit at the local speed of sound. It is surprising
that the fundamental longitudinal mode does not compare with the rotational speed since the
fan is six bladed. (The six bladed fan frequency should lead to 44 Hz).The 63 Hz frequency
seems to correspond to the fundamental radial acoustic mode based on the geometric mean
diameter between the fan and the stilling section screen. However, we lack data to positively
identify this mode.

Tunnel Wall Effects

The flow around a model in a wind tunnel differs from that in free air due effects can be
described in three categories; horizontal buoyancy, flow blocking, and trailing vortex
constraints. The effects of the first two categories are estimated when the model is at zero
degree angle of attack.

Horizontal buoyancy results from a pressure gradient along the flow direction due to tunnel
geometry and boundary layer growth, are self induced. These effects are negligible in the
Wright Brothers Wind Tunnel as long as the desired accuracy for an aerodynamic coefficient is
roughly one-half percent or larger.

Flow blockage causes in change in the effective tunnel speed. Solid blockage is a
consequence of the volume displacement of the model, and its distribution. Wake blockage is
a consequence of the displacement of fluid by the non -uniform flow in the wake. The static
pressure reference used to determine tunnel speed is located approximately at the plane of
maximum model area cross section, rather than upstream as is common. Thus the tunnel
speed setting includes the blockage corrections to a good approximation.

The trailing vorticity from a lifting model is constrained from following its desired trajectory by
the image downwash caused by tunnel floor and walls. The correction for this effect is based
upon a calculation of the downwash from image vorticity under the assumption the lift is
correct, but the angle of attack is slightly small. The side wash is assumed to be negligible.

Thus,
S
Δα = ∂ (c L )(57.3)
C 
and
S
Δc D = ∂ C L2
C 
where C = 55.00 square feet

10
 S = model wing area

The value of delta can be estimated from Figure 10.3, p385 of Reference 2; with lambda = 0.7
and k = model span/10 feet. For more details see Barlow, Ref 2, Chapter 10 sections 7 and 8.

References

1. Genn, S; Flow Behavior in the Wright Brothers Facility; Report WBWT – TR=1187 MIT
Sept 1983.
2. Barlow, Jewel, Rae, William H. jr., and Pope, Alan; Low Speed Wind Tunnel Testing,
John Wiley, New York, 1999
3. Morales, Craigh N., Smith, David B. and Walton, Joshua F.: Internal Flow Applications
of Vortex Generators, Final Report, Subject 16.622 (MIT) Spring 2004
4. D'Angelo, Michael and Durgin, Frank H,; The Measurement of Static and Dynamic Flow
Angles; with and without an upstream wing; Report WBWT – TR- 1243; Dec 1987
5. Lazar, Drahumir, Response of Pressure Probes to Longitudinal and Lateral Turbulent
Fluctuations in Isotropic Turbulence; MIT SM Thesis submitted to the Department of
Aeronautics and Astronautics, June 1979.

11
 FIGURE 1
WRIGHT BROTHERS WIND TUNNEL

12
13
14
15
 Fractional deviation of q/q Standard deviation of q (psf)

16
17
18
19
20